276 lines
11 KiB
Python
276 lines
11 KiB
Python
# 导入必要的系统库
|
||
import os
|
||
import sys
|
||
|
||
# 设置包名和添加项目根目录到系统路径
|
||
__package__ = "trainer"
|
||
sys.path.append(os.path.abspath(os.path.join(os.path.dirname(__file__), '..'))) # 将项目根目录添加到系统路径
|
||
|
||
# 导入训练所需的库
|
||
import argparse
|
||
import time
|
||
import math
|
||
import warnings
|
||
import torch
|
||
import torch.distributed as dist
|
||
from contextlib import nullcontext
|
||
from torch import optim, nn
|
||
from torch.nn.parallel import DistributedDataParallel
|
||
from torch.utils.data import DataLoader, DistributedSampler
|
||
from transformers import AutoTokenizer, AutoModelForCausalLM
|
||
from model.model_minimind import MiniMindConfig, MiniMindForCausalLM
|
||
from dataset.lm_dataset import SFTDataset
|
||
|
||
warnings.filterwarnings('ignore') # 忽略警告信息
|
||
|
||
|
||
# 日志打印函数
|
||
# 在分布式训练时只在主进程(rank=0)上打印日志
|
||
def Logger(content):
|
||
if not ddp or dist.get_rank() == 0:
|
||
print(content)
|
||
|
||
|
||
# 余弦学习率调度器
|
||
# 在训练过程中逐渐降低学习率，最终降到初始值的1/10
|
||
def get_lr(current_step, total_steps, lr):
|
||
return lr / 10 + 0.5 * lr * (1 + math.cos(math.pi * current_step / total_steps))
|
||
|
||
|
||
# 训练一个epoch的函数
|
||
def train_epoch(epoch, wandb):
|
||
# 使用交叉熵损失函数，reduction='none'以便后续通过mask处理填充token
|
||
loss_fct = nn.CrossEntropyLoss(reduction='none')
|
||
start_time = time.time()
|
||
|
||
# 遍历数据加载器获取批次数据
|
||
for step, (X, Y, loss_mask) in enumerate(train_loader):
|
||
# 将数据移动到指定设备
|
||
X = X.to(args.device) # 输入序列
|
||
Y = Y.to(args.device) # 目标序列
|
||
loss_mask = loss_mask.to(args.device) # 损失掩码
|
||
|
||
# 更新学习率
|
||
lr = get_lr(epoch * iter_per_epoch + step, args.epochs * iter_per_epoch, args.learning_rate)
|
||
for param_group in optimizer.param_groups:
|
||
param_group['lr'] = lr
|
||
|
||
# 前向传播和损失计算
|
||
with ctx: # 使用自动混合精度训练
|
||
res = model(X) # 模型前向传播
|
||
# 计算交叉熵损失
|
||
loss = loss_fct(
|
||
res.logits.view(-1, res.logits.size(-1)), # 展平logits
|
||
Y.view(-1) # 展平目标序列
|
||
).view(Y.size())
|
||
|
||
# 应用损失掩码并计算平均损失
|
||
loss = (loss * loss_mask).sum() / loss_mask.sum()
|
||
loss += res.aux_loss # 添加辅助损失（如果有的话，例如MoE的负载均衡损失）
|
||
loss = loss / args.accumulation_steps # 根据梯度累积步数缩放损失
|
||
|
||
# 反向传播
|
||
scaler.scale(loss).backward() # 使用梯度缩放器缩放损失并反向传播
|
||
|
||
# 梯度更新
|
||
if (step + 1) % args.accumulation_steps == 0:
|
||
scaler.unscale_(optimizer) # 反缩放梯度
|
||
# 梯度裁剪，防止梯度爆炸
|
||
torch.nn.utils.clip_grad_norm_(model.parameters(), args.grad_clip)
|
||
|
||
scaler.step(optimizer) # 优化器更新参数
|
||
scaler.update() # 更新梯度缩放器
|
||
|
||
optimizer.zero_grad(set_to_none=True) # 清空梯度
|
||
|
||
# 打印训练日志
|
||
if step % args.log_interval == 0:
|
||
spend_time = time.time() - start_time
|
||
Logger(
|
||
'Epoch:[{}/{}]({}/{}) loss:{:.3f} lr:{:.12f} epoch_Time:{}min:'.format(
|
||
epoch + 1,
|
||
args.epochs,
|
||
step,
|
||
iter_per_epoch,
|
||
loss.item() * args.accumulation_steps,
|
||
optimizer.param_groups[-1]['lr'],
|
||
spend_time / (step + 1) * iter_per_epoch // 60 - spend_time // 60))
|
||
|
||
# 记录wandb日志（如果启用）
|
||
if (wandb is not None) and (not ddp or dist.get_rank() == 0):
|
||
wandb.log({"loss": loss * args.accumulation_steps,
|
||
"lr": optimizer.param_groups[-1]['lr'],
|
||
"epoch_Time": spend_time / (step + 1) * iter_per_epoch // 60 - spend_time // 60})
|
||
|
||
# 保存模型检查点
|
||
if (step + 1) % args.save_interval == 0 and (not ddp or dist.get_rank() == 0):
|
||
model.eval() # 切换到评估模式
|
||
# 根据是否使用MoE设置检查点路径
|
||
moe_path = '_moe' if lm_config.use_moe else ''
|
||
ckp = f'{args.save_dir}/full_sft_{lm_config.hidden_size}{moe_path}.pth'
|
||
|
||
# 获取模型状态字典
|
||
if isinstance(model, torch.nn.parallel.DistributedDataParallel):
|
||
state_dict = model.module.state_dict()
|
||
else:
|
||
state_dict = model.state_dict()
|
||
|
||
# 将模型参数转换为半精度格式并保存
|
||
state_dict = {k: v.half() for k, v in state_dict.items()}
|
||
torch.save(state_dict, ckp)
|
||
model.train() # 切换回训练模式
|
||
|
||
|
||
# 初始化模型和分词器
|
||
def init_model(lm_config):
|
||
# 加载预训练的分词器
|
||
tokenizer = AutoTokenizer.from_pretrained('../model')
|
||
# 初始化因果语言模型
|
||
model = MiniMindForCausalLM(lm_config)
|
||
# 根据是否使用MoE设置权重路径
|
||
moe_path = '_moe' if lm_config.use_moe else ''
|
||
ckp = f'{args.save_dir}/pretrain_{lm_config.hidden_size}{moe_path}.pth'
|
||
# 加载预训练权重
|
||
state_dict = torch.load(ckp, map_location=args.device)
|
||
model.load_state_dict(state_dict, strict=False)
|
||
|
||
# 打印模型可训练参数总量
|
||
Logger(f'LLM可训练总参数量：{sum(p.numel() for p in model.parameters() if p.requires_grad) / 1e6:.3f} 百万')
|
||
# 将模型移动到指定设备
|
||
model = model.to(args.device)
|
||
return model, tokenizer
|
||
|
||
|
||
# 初始化分布式训练环境
|
||
def init_distributed_mode():
|
||
if not ddp: return
|
||
global ddp_local_rank, DEVICE
|
||
|
||
# 初始化分布式进程组
|
||
dist.init_process_group(backend="nccl")
|
||
# 获取分布式训练的相关参数
|
||
ddp_rank = int(os.environ["RANK"]) # 全局进程编号
|
||
ddp_local_rank = int(os.environ["LOCAL_RANK"]) # 本地进程编号
|
||
ddp_world_size = int(os.environ["WORLD_SIZE"]) # 总进程数
|
||
# 设置当前进程使用的GPU
|
||
DEVICE = f"cuda:{ddp_local_rank}"
|
||
torch.cuda.set_device(DEVICE)
|
||
|
||
|
||
# 主函数
|
||
if __name__ == "__main__":
|
||
# 配置命令行参数
|
||
parser = argparse.ArgumentParser(description="MiniMind Full SFT")
|
||
# 基础训练参数
|
||
parser.add_argument("--out_dir", type=str, default="../out", help="输出目录")
|
||
parser.add_argument("--epochs", type=int, default=2, help="训练轮数")
|
||
parser.add_argument("--batch_size", type=int, default=16, help="批次大小")
|
||
parser.add_argument("--learning_rate", type=float, default=5e-7, help="学习率")
|
||
parser.add_argument("--device", type=str, default="cuda:0" if torch.cuda.is_available() else "cpu", help="训练设备")
|
||
parser.add_argument("--dtype", type=str, default="bfloat16", help="训练精度")
|
||
|
||
# 日志和监控参数
|
||
parser.add_argument("--use_wandb", action="store_true", help="是否使用wandb记录训练过程")
|
||
parser.add_argument("--wandb_project", type=str, default="MiniMind-Full-SFT", help="wandb项目名称")
|
||
parser.add_argument("--log_interval", type=int, default=100, help="日志打印间隔")
|
||
parser.add_argument("--save_interval", type=int, default=100, help="模型保存间隔")
|
||
|
||
# 分布式训练参数
|
||
parser.add_argument("--num_workers", type=int, default=1, help="数据加载进程数")
|
||
parser.add_argument("--ddp", action="store_true", help="是否使用分布式训练")
|
||
parser.add_argument('--local_rank', type=int, default=-1, help="分布式训练的本地进程编号")
|
||
|
||
# 优化器参数
|
||
parser.add_argument("--accumulation_steps", type=int, default=1, help="梯度累积步数")
|
||
parser.add_argument("--grad_clip", type=float, default=1.0, help="梯度裁剪阈值")
|
||
parser.add_argument("--warmup_iters", type=int, default=0, help="学习率预热迭代次数")
|
||
|
||
# 模型参数
|
||
parser.add_argument('--hidden_size', default=512, type=int, help="隐藏层维度")
|
||
parser.add_argument('--num_hidden_layers', default=8, type=int, help="Transformer层数")
|
||
parser.add_argument('--max_seq_len', default=512, type=int, help="最大序列长度")
|
||
parser.add_argument('--use_moe', default=False, type=bool, help="是否使用MoE")
|
||
parser.add_argument("--data_path", type=str, default="../dataset/sft_mini_512.jsonl", help="训练数据路径")
|
||
parser.add_argument("--wandb_api_key", type=str, default=None, help="WandB API Key，用于无交互环境自动登录")
|
||
|
||
args = parser.parse_args()
|
||
|
||
# 初始化模型配置
|
||
lm_config = MiniMindConfig(hidden_size=args.hidden_size, num_hidden_layers=args.num_hidden_layers,
|
||
use_moe=args.use_moe)
|
||
# 创建输出目录
|
||
args.save_dir = os.path.join(args.out_dir)
|
||
os.makedirs(args.save_dir, exist_ok=True)
|
||
os.makedirs(args.out_dir, exist_ok=True)
|
||
|
||
# 计算每次迭代处理的token数量
|
||
tokens_per_iter = args.batch_size * args.max_seq_len
|
||
device_type = "cuda" if "cuda" in args.device else "cpu"
|
||
|
||
# 设置wandb运行名称
|
||
args.wandb_run_name = f"MiniMind-Full-SFT-Epoch-{args.epochs}-BatchSize-{args.batch_size}-LearningRate-{args.learning_rate}"
|
||
|
||
# 设置自动混合精度训练上下文
|
||
ctx = nullcontext() if device_type == "cpu" else torch.cuda.amp.autocast()
|
||
# 检查是否为分布式训练
|
||
ddp = int(os.environ.get("RANK", -1)) != -1
|
||
ddp_local_rank, DEVICE = 0, "cuda:0"
|
||
|
||
# 设置随机种子
|
||
base_seed = 1337
|
||
torch.manual_seed(base_seed)
|
||
torch.cuda.manual_seed(base_seed)
|
||
|
||
# 初始化分布式训练环境
|
||
if ddp:
|
||
init_distributed_mode()
|
||
args.device = torch.device(DEVICE)
|
||
rank = dist.get_rank()
|
||
torch.manual_seed(base_seed + rank)
|
||
torch.cuda.manual_seed(base_seed + rank)
|
||
|
||
# 初始化wandb（如果启用）
|
||
if args.use_wandb and (not ddp or ddp_local_rank == 0):
|
||
import wandb
|
||
# 优先使用命令行参数，其次环境变量
|
||
api_key = args.wandb_api_key or os.environ.get("WANDB_API_KEY", None)
|
||
if api_key is not None:
|
||
# 自动登录，适用于无交互环境
|
||
wandb.login(key=api_key)
|
||
# 初始化 wandb 任务
|
||
wandb.init(project=args.wandb_project, name=args.wandb_run_name)
|
||
else:
|
||
wandb = None
|
||
|
||
# 初始化模型和分词器
|
||
model, tokenizer = init_model(lm_config)
|
||
|
||
# 准备训练数据
|
||
train_ds = SFTDataset(args.data_path, tokenizer, max_length=args.max_seq_len)
|
||
train_sampler = DistributedSampler(train_ds) if ddp else None
|
||
train_loader = DataLoader(
|
||
train_ds,
|
||
batch_size=args.batch_size,
|
||
pin_memory=True,
|
||
drop_last=False,
|
||
shuffle=False,
|
||
num_workers=args.num_workers,
|
||
sampler=train_sampler
|
||
)
|
||
|
||
# 初始化优化器和梯度缩放器
|
||
scaler = torch.cuda.amp.GradScaler(enabled=(args.dtype in ['float16', 'bfloat16']))
|
||
optimizer = optim.AdamW(model.parameters(), lr=args.learning_rate)
|
||
|
||
# 配置分布式训练
|
||
if ddp:
|
||
model._ddp_params_and_buffers_to_ignore = {"pos_cis"}
|
||
model = DistributedDataParallel(model, device_ids=[ddp_local_rank])
|
||
|
||
# 计算每个epoch的迭代次数
|
||
iter_per_epoch = len(train_loader)
|
||
|
||
# 开始训练循环
|
||
for epoch in range(args.epochs):
|
||
train_epoch(epoch, wandb)
|